Cyber Security on the Automatic Generation Control

Abstract

Due to the advancement of communication technologies and the integration of power systems with communication infrastructure, there has been a significant increase in the susceptibility to cyber attacks targeting the operation of power systems. The rising cyber vulnerabilities raise multiple concerns related to the stability and reliability of power systems, particularly in light of major blackouts that occurred in North America and some European countries. Automatic Generation Control (AGC) is a solution to maintain the frequency of power systems within a predefined range by effectively addressing any frequency deviation out of the permissible range, and consequently ensuring power system stability. Though, the AGC system is susceptible to cyber attacks as it obtains the required data through communication links. Therefore, designing a robust protection mechanism for such systems is crucial in order to detect, allocate, and mitigate cyber attacks effectively. In this thesis, different data-driven architectures are proposed for detecting, allocating, and mitigating False Data Injection Attacks (FDIA) against the AGC systems considering the nonlinearities. The first proposed model is a detection model with a remarkable capability to detect and allocate both individual and coordinated stealthy cyber attacks, which pose a substantial threat to the effectiveness of the detection systems. The second and third models are comprehensive mitigation system (CMS) and attacked class-based mitigation model (ACM), which are utilized to precisely recover the attacked measurements. The proposed models undergo extensive training and testing using a wide range of scenarios involving Pulse and Ramp stealthy attacks to ensure their effectiveness in both steady-state and transient conditions. Furthermore, the integration of photovoltaic systems into two-areas AGC systems is put forward in order to study the impact of PV integration on the proposed detection system, marking a pioneering initiative in this thesis. The proposed model is evaluated using two-areas AGC system, and the results obtained from the evaluation showcase a remarkable demonstration of superior performance.

Date of Award	14 Nov 2023
Original language	American English
Supervisor	Ameena Alsumaiti (Supervisor)